Experimental efficiency calibration for a HPGe gamma-ray ... · Experimental efficiency calibration for a HPGe gamma-ray detector NKS GammaSpec 2017 Seminar DTU Nutech, Roskilde,

SÄTEILYTURVAKESKUS • STRÅLSÄKERHETSCENTRALEN

RADIATION AND NUCLEAR SAFETY AUTHORITY

Experimental efficiency calibration for a HPGe gamma-ray detector

NKS GammaSpec 2017 Seminar

DTU Nutech, Roskilde, Denmark

19-20.9.2017

Roy Pöllänen Environmental Radiation Surveillance and Emergency Preparedness


RADIATION AND NUCLEAR SAFETY AUTHORITY 19-20.9.2017

NKS - GammaSpec 2017

2

Outline

1. Introduction (STUK’s -ray lab)

2. The problem … and how to overcome it

3. Peak efficiency

4. Total efficiency

5. Methodology of the uncertainty determination

6. Final result, comparison between uncertainties and conclusions




3

• 16 HPGe spectrometers (7 Ortec, 9 Canberra).

• 4 electrically cooled, others are LN2-cooled (5 Möbius, 7 Cryo-Cycle).

• Digital MCAs (different DSPEC generations).

• 3000 4000 analyses per year.

1 2

1. Introduction (STUK’s -ray lab)




4

Spectrum analysis

• GAMMA-99 software developed by STUK – cascade summing correction since 1983 !

– sample height and density corrections

– automatic and interactive operation modes




5

• UniSampo/Shaman

– UniSampo: peak search, baseline & peak fitting, peak areas

– Shaman: rule-based peak identification and activity calculation, mimics human analyst

Interactive fitting




6

Measurement geometries

Simple cylindrical:

• 0-30 mL, free sample height

• 0-100 mL, free sample height

Marinelli:

• 0.5 L, fixed sample height

Some special geometries, too

Marinelli T-beaker

W-beaker




7

Calibrations & the NAMIT software

Detectors (here 5 different) Energy Peak Eff. Tot. Eff. Resolution




8

• Present efficiency calibrations are not adequately traceable.

• In general, the calibrations are ok, but some deficiencies have been detected (for example: in some cases there have been shortcomings in small energies).

• Unclear and non-traceable uncertainty calculation.

2. The problem …




9

Different measurement geometries and different detector characteristics

– Typically 25 measurement geometries/detector – 16 detectors – 1 FEP efficiency curve + 1 Tot. efficiency curve per geometry

Total number of the efficiency curves: ∿3 × 16 × 2 ≈ 100 !

The ways of performing efficiency determination properly must be designed carefully.




10

• Different type of samples must be measured:

– Air filters. Not a very complex case (from the efficiency calibration point of view).

– General environmental samples. Measurements done using predetermined geometries. Challenge: various sample mass, density and elemental composition.

– Other samples. Other than ”normal” geometries possible. Efficiency calculated by computational means (DECCA, EFFTRAN, VGSL, GESPECOR,…).




11

• Existing calibrations have previously been done as follows: – Certified cylindrical source of X mm in thickness (typically X = 5 mm)

was measured first.

– Using an appropriate detector model the calibrations were computed by VGSL for the source thickness of 0 mm.

– Real sample density and height are corrected by the spectrum analysis software.

Model preparation & tuning

Certified reference source

26 mm

VGSL model for the certified source

Model for the source thickness

of 0 mm

Peak efficiency computed

for the source thickness

of 0 mm

Analysis software takes into account

sample h & ρ

Sample to be analyzed

h ρ




12

• Certified sources of thickness 0 mm were purchased → no need to recalculate/correct the efficiencies back to the thickness of 0 mm.

• Minimizing the use of MC-methods or other similar type of computations (but their use cannot be avoided).

• ∿ 100 efficiency curves → an individual specialist cannot process all these → the pain of making efficiency calibrations must be shared between the analysts →

advantage: rising the expertise.

• Special attention to the uncertainty calculation.

... and how to overcome it



Number of detected gammas per unit time

Number of emitted gammas per unit time

19-20.9.2017


13

Peak efficiency, i , for the gammas of energy i

i

3. Peak efficiency

This is obtained

by measurements

(with appropriate

corrections)

This is reported

in the source

certificate




14

• In the following, a BeGe detector labelled as B6 and a W-beaker with close measurement geometry is considered.

• A certified reference source of thickness ∿ 0 was measured (dead time 3.6% → no significant random coincidences).

• The source contained nuclides 210Pb … 88Y ( 46 … 1836 keV).

210Pb

241Am

109Cd

57Co

139Ce 113Sn 113Sn 51Cr 85Sr 137Cs



0.01

0.10

1.00

10 100 1000 10000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

Full Energy Peak Efficiency (FEPE) for detector B6

Existing FEPE

Full Energy Peak Efficiency for the detector B6

19-20.9.2017


15

• The very first result



0.01

0.10

1.00

10 100 1000 10000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

Full Energy Peak Efficiency (FEPE) for detector B6

Existing FEPE


19-20.9.2017


16

210Pb 241Am 109Cd

51Cr

137Cs

113Sn (IT)

54Mn

65Zn

Nuclide for

which the CC

is 1 (or ∿1)

• The very first result




17

• Secondly, Unisampo-Shaman true coincidence correction factors were taken into account (+ preliminary uncertainties).

Much better result, but still some deficiencies. Why?

0.01

0.10

1.00

10 100 1000

Pe

k ef

fice

ncy

(cp

s/B

q)

Energy (keV)

Peak efficiency for the detector B6

Existing FEPE




18

• I contacted with the developer of the Unisampo-Shaman software → several calibration nuclides have - X-ray true coincidences, but there were deficiencies in the decay scheme data library (X-ray cutoff 15 keV, X-ray energies lumped).

Note: In the

case of the

BeGe detectors

the FEPE

values are not

negligible for

low energies.




19

• Almost perfect FEPE values! However, there are still points at the energy of 166 keV (139Ce) and 255 keV (113Sn) that are out of the FEPE curve. In addition, low-energy (<46 keV) values are missing.

Total efficiencies must be checked!

0.01

0.10

1.00

10 100 1000 10000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)


Existing FEPE

139Ce 113Sn




20

• Problem in the case of low-energy ’s: appropriate nuclides (and having no CC) are difficult to find → let’s try to use X-rays!

• Another problem: In terms of energy the X-rays are located close to each other → the peaks are lumped → Gamma99 software was used for the lumped peak areas!

• Example: 210Pb.

9.6%

11%

2.6%




21

210Pb (46.5 keV)

Ge escape

Background

Bi L Bi L

Bi L

Rough efficiency estimate from the X-ray peaks can be obtained.

10000

1000

100

10

1




22

• 137Cs (Ba) x-rays (32 and 37 keV) and 241Am 26-keV gamma were also taken into account.

0.01

0.10

1.00

10 100 1000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

241Am 26 keV

137Cs X-rays

210Pb X-rays

241Am 26 keV

Quite nice! But still: 139Ce and 113Sn 255 keV should fit better.

!




23

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

10 100 1000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

Linear y-axis:




24

4. Total efficiency

• tot is used for the determination of the true coincidence correction.

• tot must be redetermined in order to get better ”fit” for the peak efficiency of 139Ce (166 keV) and 113Sn (255 keV).

• VGSL2 and GESPECOR calculations were performed to fix the input data → same input models were used for simulating tot → new coincidence correction values for the determination of the peak efficiency.




25

MC-simulated peak efficiencies (and the DECCA result)

Not a bad result!

0.00

0.10

0.20

0.30

0.40

0.50

10.0 100.0 1000.0

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

VGSL2

GESPECOR

DECCA




26

Present and MC-simulated total efficiencies

Notable differences between the present and simulated total

efficiencies! Average value of VGSL and GESPECOR were

selected (except for the energies of 10 keV and 3.6 MeV)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

10.000 100.000 1000.000

Tota

l eff

icie

ncy

(cp

s/B

q)

Energia (keV)

Present value

VGSL2

GESPECOR

210Pb

137Cs




27




28

5. Methodology of the uncertainty

• Efficiency values are out of relevance if there is no proper uncertainty estimation → efficiency uncertainties are part of the full uncertainty budget when “real” samples are under investigation.

• Components affecting the uncertainty

– must be identified and

– must be quantified

• Components not included in the uncertainty calculation must be identified (here e.g. inhomogeneity of the ”real” samples and gamma-ray attenuation).

determination




29

Note! Present practice/convention in the uncertainty estimation of the efficiencies: Uncertainties associated with the samples under investigation are accounted for in the peak efficiency uncertainty determination !!!???




30

Standard uncertainty propagation formula used here

The uncertainties are not only associated with ni, Ai or I. The corrections also have uncertainties:

– Decay correction prior to the counting period

– Decay correction during the counting period

– Random coincidence summing correction

– True coincidence summing correction

– Correction for the self attenuation

– Correction for the source inhomogeneity

– Correction for other source characteristics

– …




31

A. Equipment used in the measurements

B. Measurement conditions (environment)

C. Characteristics of the reference source

D. Measurement geometry

E. Data acquisition

F. Spectrum analysis

G. Other points influencing the uncertainties

At least following points/items must be accounted for to determine efficiency uncertainties:




32

A. Equipment used in the measurements

• Stability especially when long data acquisition times are used (variation of the resolution, energy and efficiency calibration in time.

• MCA clock and settings (conversion gain, pole zero, pulse shaping time,…).

Not estimated but most probably their contribution to the uncertainty << 1%




33

B. Measurement conditions (environment)

• Variation of the background as a function of time (routine background measurements and routine Rn monitoring).

• Means to prevent contamination.

• Variation of the air temperature, pressure and moisture (air conditioning).

Not quantified but certainly << 1%




34

C. Characteristics of the reference sources

• Testing the source (filter material) homogeneity in NPL

(k=2)

Text in the source certificate:




35

Imaging plate picture

Beaker inner diameter 42 mm

Ø 42 mm

Does inhomogenous

radionuclide distribution

have an effect?




36

• Correction for the source diameter (42 mm → 40 mm): – Without correction, smaller reference source diameter lead to

slightly higher peak efficiency (but probably a small effect).

– DECCA-calculations of the efficiency was performed by assuming smaller diameters.

– Fitting the DECCA-results using a ”smooth” function allows applying the correction to the energies of the reference nuclides.

1

1.002

1.004

1.006

1.008

1.01

1.012

10 100 1000 10000

Pea

k ef

fici

ency

rat

io

Energy (keV)

40mm/42mm

38mm/42mm Effect < 1%

No other programs (such as MCNP) were used to quantify the effect of inhomogeneous radionuclide distribution!




37

• Correction for the thickness of the reference source: – Whatmann GF 10 filter paper of thickness 0.35 mm and density

0.2 g cm-3.

– Without correction, reference source thickness lead to slightly lower peak efficiency.

– DECCA program and fitting → less than 0.5% effect.

Note! the effects of true diameter and thickness of the reference source were accounted for as corrections (having no uncertainty). Additional uncertainty of 1% (not 2.5% as suggested by NPL) was used for the peak efficiencies.




38

D. Measurement geometry

At least following points must be quantified:

• Accuracy of the positioning of the reference source with respect to the symmetry axis of the detector ( ± 1 mm assumed, no sample guide used).

• In the case of ”real” samples the uncertainty of the positioning might be up to ± 5 mm.

• Dimension, thickness, density and composition variation of the beaker (container walls and bottom)




39

Example: The effect of the source horizontal displacement

Effect < 1%

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 500 1000 1500 2000 2500 3000

Effi

cie

ncy

dif

fere

nce

(%

)

Energy (keV)

Deviation 5 mm

Difference of the efficiency (%) compared to the case when the source is in the symmetry axis (VGSL computation)

Deviation 1 mm




40

E. Data acquisition

• Dead time in the reference source measurements < 5% → no random coincidences.

• Data acquisition time selected so that the areas of the main peaks ≈ 105 → statistical uncertainty < 1%.



0.97

0.98

0.99

1

1.01

1.02

1.03

30.0 300.0 3000.0Energy (keV)

Count rate difference (USS vs. G99 software)

19-20.9.2017


41

57Co (122 keV)

> 105 counts!

F. Spectrum analysis

UniSampo-Shaman and Gamma99 software used in the analysis.

Additional uncertainty of 0.5% added to the peak areas




42

Coincidence correction (and its uncertainty) is of relevance!

COINCIDENCE CORRECTION

Nuclide t 1/2 E (keV) ccG99 ccUSS ccEFFTRAN ccGESPECOR ccFinal Δcc

Pb-210 22.23(12) a 46.5 1 1 1 1 1.000 0.000

Am-241 432.6(6) a 59.5 1 1 1.022 1 1.000 0.011

Cd-109 461.9(4) d 88.0 1 1 1 1 1.000 0.000

Co-57 271.80(5) d 122.1 1.02 1.058 1.11 1.1013 1.039 0.042

Co-57 271.80(5) d 136.4 0.86 0.87 0.816 0.9129 0.865 0.040

Ce-139 137.641(20) d 165.9 1.43 1.441 1.355 1.3227 1.387 0.058

Sn-113EC 115.09(3) d 255.2 1.35 1.347 1.401 1.352 1.349 0.026

Cr-51 27.704(4) d 320.1 0.98 1.009 1 1.0086 0.995 0.014

Sn-113IT 115.09(3) d 391.7 1 1 1 1 1.000 0.000

Sr-85 64.850(7) d 514.0 1.14 1.12 1.223 1.097 1.130 0.055

Cs-137 30.05(8) a 661.7 1 1 1 1 1.000 0.000

Mn-54 312.19(3) d 834.8 0.99 1.012 1.024 1.0155 1.001 0.014

Y-88 106.63(5) d 898.0 1.43 1.403 1.569 1.4652 1.508 0.073

Zn-65 244.01(9) d 1115.5 1.02 1.034 1.067 1.0866 1.027 0.030

Co-60 5.2711(8) a 1173.2 1.27 1.269 1.293 1.2656 1.270 0.013

Co-60 5.2711(8) a 1332.5 1.28 1.283 1.307 1.2776 1.282 0.014

Y-88 106.63(5) d 1836.1 1.49 1.457 1.634 1.5213 1.474 0.077

∿5% unc.




43

G. Other points influencing the uncertainties

• Elemental composition of the sample to be analyzed.

• Uncertainties associated with the measurement of the sample height (± 1 mm assumed) and density (mass determined with hich accuracy) determination.

• Variation of the beaker wall thickness (± 0.1 mm assumed); because of geometrical reasons this is not relevant for the reference source of thickness ∿0 mm.

• Horizontal displacement of the sample (± 5 mm assumed).




44

DECCA computation: The incluence of the elemental composition to the peak efficiency, sample height 26 mm.

0

0.1

0.2

0.3

0.4

10 100 1000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energy (keV)

W0-calibration

H2O, basic case.

Air

MgSiOPE

GRA




45

DECCA computation: The incluence of the elemental composition to the peak efficiency compared with the case of H2O (density = 1 for all cases). Sample height 26 mm.

-8

-6

-4

-2

0

2

4

6

8

10 100 1000 10000

Dif

fere

nce

(%

)

GRA

PE

SiO

Mg

Air

Conservative (?) selection for the uncertainty estimation: GRA




46

DECCA computation: The incluence of the sample height uncertainty (here ± 1 mm assumed for a sample of 20 mm in thickness; sample densities 1 and 0.3).

The effect is a bit larger in the case of a 5 mm thick source. This is accounted for in the uncertainty estimation.

0

0.1

0.2

0.3

0.4

10 100 1000

Pe

ak e

ffic

ien

cy (

cps/

Bq

)

Energia (keV)

Sample heights 19, 20 and 21 mm

ρ = 1

ρ = 0.3

W0-calibration




47

5. Final result, comparison between the

uncertainties and conclusions

0.01

0.10

1.00

10.0 100.0 1000.0

Effi

cie

ncy

Energy (keV)

Total efficiency

Peak efficiency

E (keV) (cps/Bq) Unc. (%)

10.8 0.127 49.1

13.0 0.151 46.1

15.6 0.198 42.1

26.3 0.287 25.1

32.0 0.312 17.2

36.8 0.332 10.3

46.5 0.339 6.9

59.5 0.346 5.6

88.0 0.348 4.9

122.1 0.333 5.7

136.4 0.318 6.2

165.9 0.295 5.8

255.2 0.212 4.9

320.1 0.162 3.9

391.7 0.133 3.6

514.0 0.103 6.0

661.7 0.081 3.5

834.8 0.066 3.7

898.0 0.063 5.9

1115.5 0.051 4.6

1173.2 0.048 3.6

1332.5 0.043 3.6

1836.1 0.032 6.2

3600.0 0.016 9.9




48

0.01%

0.10%

1.00%

10.00%

30.0 300.0 3000.0

Un

cert

ain

ty

Energy (keV)

Combined Uncertainty (k=1)

Coincidence Corr. Uncertainty

Sample height +-1mm

Sample elemental composition

Container wall thickness +-0.1 mm

Gamma-ray yield uncertainty

Container radial displacement +-5mm

USS count rate uncertainty

USS, cps additional unc.

Container bottom +-0.1mm

Ref.source inhomogeneity

Ref.source displacement +-1 mm

Uncertainty components as a function of energy (46 keV – 1836 keV)




49

Main uncertainty components: Coincidence correction, elemental composition and sample height of the ”real” sample.

0.10%

1.00%

10.00%

30.0 300.0 3000.0

Un

cert

ain

ty

Energy (keV)

Combined Uncertainty (k=1)

Coincidence Corr. Uncertainty

Sample height +-1mm

Sample elemental composition

Container wall thickness +-0.1 mm

Gamma-ray yield uncertainty

Container radial displacement +-5mm




50

A full and traceable efficiency determination with thorough uncertainty estimation is a tedious process including measurements and simulations/computations.

How to apply the methodology for a large number of detectors of different characteristics?

Main uncertainties are attached to the sample under investigation

Uncertainties associated with the efficiency calibration and those of the samples must be separated.

The results must also be compared in the case of other certified reference sources.

Documents

Experimental efficiency calibration for a HPGe gamma-ray ... · Experimental efficiency calibration for a HPGe gamma-ray detector NKS GammaSpec 2017 Seminar DTU Nutech, Roskilde,